Pith. sign in

REVIEW 6 major objections 6 minor 47 references

Scalable Entanglement Detection in Quantum Systems via Fisher Linear Discriminant Analysis

T0 review · 6 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that a classical linear classifier, trained on Pauli expectation values, separates entangled from separable two-, three-, and four-qubit states with high accuracy in noise-free simulations.

desk verdict A clean but routine FLDA demo on small simulated states; the 'scalable' claim is contradicted by the method's own exponential feature-space cost. read the letter →

arxiv 2509.03233 v2 pith:MESPK74S submitted 2025-09-03 quant-ph

Mahmoud Mahdian , Zahra Mousavi This is my paper · ORCID
classification quant-ph
keywords quantumentanglementdetectionFisherlineardiscriminantanalysisWernerstatespositivepartialtransposemachinelearningstateclassificationPauliobservables
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that Fisher linear discriminant analysis (FLDA), a classical statistical classifier, can detect quantum entanglement by learning a single linear projection of measured Pauli expectation values. In noise-free simulations on two-, three-, and four-qubit systems it reports high test accuracy for Werner states, positive-partial-transpose states, and a biseparable family, with accuracy degrading as the entangled and separable classes overlap more heavily. The largest drop appears in the two-qubit concurrence-based family, where high-overlap test accuracy is reported as 0.33. The motivation is practical: FLDA reduces the exponentially large space of Pauli observables to a one-dimensional score and returns a discriminant vector that identifies which measurements carry the most entanglement information. The paper states explicitly that noise was not included in its idealized simulations, so the accuracies describe an ideal measurement limit rather than an immediate experimental method.

What carries the argument

The central object is the Fisher linear discriminant vector $\mathbf{w}$, obtained by maximizing the Fisher criterion $J(\mathbf{w}) = \frac{\mathbf{w}^{T} S_B \mathbf{w}}{\mathbf{w}^{T} S_W \mathbf{w}}$ and solving the generalized eigenvalue problem $S_B \mathbf{w} = \lambda S_W \mathbf{w}$, where $S_B$ is the between-class scatter and $S_W$ the within-class scatter of the feature vectors. The feature vectors are built from measurement outcomes, specifically expectation values $\langle \sigma_k \rangle = \mathrm{tr}(\rho \sigma_k)$ of single- and multi-qubit Pauli observables, so the method is a direct pipeline from quantum measurements to a classical one-dimensional score. The labeled examples are Werner states with mixing parameter $p$, PPT entangled states, and biseparable states, with separability assigned by the Peres-Horodecki criterion or by the construction of the state. When $S_W$ is singular it is regularized as $S_W + \epsilon I$, and classification is done by comparing the projected test point to the projected class means.

What would settle it

Re-run the two-, three-, and four-qubit experiments with the same state families and FLDA pipeline, but estimate each Pauli expectation value from a finite number of projective measurements (for example, 1000 shots) rather than from the exact trace. If the high-overlap test accuracies collapse toward chance — in particular if the two-qubit Werner 0.89 and three-qubit GHZ 0.87 figures cannot be reproduced — the noiseless-simulation claim is isolated to an ideal limit and the practical detection claim fails.

Watch

Extended reading notes

Core claim

The central claim is that a Fisher linear discriminant computed from exact Pauli expectation values separates entangled from separable states with high accuracy for the state families tested. For two-qubit Werner states the classifier reaches 0.89, 0.92, and 1.00 test accuracy in the high-, medium-, and low-overlap regimes; for the two-qubit concurrence-parametrized family the high-overlap accuracy is 0.33 while the low-overlap accuracy is 1.00. Three-qubit GHZ-type Werner states score 0.87, 0.96, and 1.00 across the same overlap levels, three-qubit PPT families score between 0.65 and 1.00, and the tested three-qubit biseparable and four-qubit Werner cases reach 1.00 even at high overlap. The paper interprets the decline with overlap as the expected behavior of a linear method: when the projected class means are close, the Fisher criterion is small and classification becomes unreliable. It also treats the discriminant vector as an interpretable output that ranks the Pauli observables by their usefulness for entanglement detection.

Load-bearing premise

The reported accuracies depend on measurement data with no noise or statistical fluctuation — the simulations use exact average outcomes — and the paper itself notes that real experimental noise would almost certainly lower them.

Editorial extensions

If this is right

  • A single scalar projection of Pauli expectation values suffices to separate entangled from separable states for the Werner, PPT, and biseparable families tested, so FLDA offers a low-cost classification route in noise-free settings.
  • The discriminant vector ranks observables by their influence on the entanglement decision, which can guide experimental measurement selection toward a small set of informative Pauli terms.
  • Class separability, as measured by the Fisher criterion, tracks classification accuracy: well-separated classes reach perfect test accuracy, while high-overlap classes degrade, most sharply for the two-qubit concurrence case at 0.33.
  • Because FLDA reduces the $4^N-1$-dimensional Pauli feature space to one dimension for a two-class problem, the per-classification cost depends on the number of measured features, not on the exponential Hilbert-space dimension.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural test the paper does not run is to inject finite shot noise by sampling Pauli outcomes according to the Born rule instead of using exact expectation values; the paper's own discussion predicts that noise inflates the within-class scatter and lowers accuracy, and the high-overlap concurrence case already sits at 0.33 in the noiseless limit.
  • The FLDA boundary is functionally an entanglement witness: the discriminant vector defines a linear functional on the space of Pauli expectation values whose sign is the classification decision, so the trained classifier could be recast as a witness whose validity for the tested families could in principle be certified rather than only measured empirically.
  • Testing on states outside the parametrized families — random mixed states, bound entangled states in other dimensions, or states whose separability boundary is nonlinear in Pauli space — would show whether the linear boundary generalizes or is specific to Werner-type structure.
  • The four-qubit perfect score at high overlap should be read as an upper bound on what the method can do, since the two- and three-qubit results show accuracy falling with overlap; realistic extensions with noise would likely require more informative features or nonlinear classifiers.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

6 major / 6 minor

Summary. The manuscript proposes to use classical Fisher Linear Discriminant Analysis (FLDA) as a machine-learning tool for classifying quantum states as entangled or separable. Feature vectors are formed from expectation values of Pauli observables; the between-class and within-class scatter matrices S_B and S_W are computed on a labeled training set (labels from the Peres-Horodecki criterion or from concurrence), and the generalized eigenproblem S_B w = λ S_W w is solved to obtain a projection direction. The authors report train/test accuracies for two-qubit Werner and concurrence states, three-qubit GHZ-based Werner, PPT, 'alternative PPT', and biseparable states, and one four-qubit Werner case, and conclude that the method is simple, efficient, scalable, interpretable, and promising for real-world experiments.

Significance. The FLDA formalism itself is standard, and applying it to Pauli-expectation-value features is a natural idea that could in principle yield an interpretable baseline for entanglement classification. However, the manuscript does not substantiate its central claims: the scalability claim is contradicted by the exponential feature-space construction, the three-qubit and 'alternative PPT' benchmarks contain physics errors, the Werner benchmark is near-tautological because the label parameter appears directly in the features, and the one non-trivial two-qubit result falls below chance in the high-overlap case. The paper also provides no code, data, hyperparameters, or error bars, and its own discussion concedes that noise was omitted. As a result, the reported accuracies do not establish a practical or scalable entanglement-detection method.

major comments (6)
  1. [Section 3.2, Eq. (11), Table 5] The state called an 'alternative PPT entangled state' in Eq. (11) is not entangled. Expanding ρ = (1/8)(I⊗I⊗I + I⊗σ_z⊗σ_z + σ_z⊗I⊗σ_z + σ_z⊗σ_z⊗I) in the computational basis gives a diagonal matrix with entries 1/2 on |000⟩ and |111⟩ and zero elsewhere, i.e., ρ = (|000⟩⟨000| + |111⟩⟨111|)/2, which is a separable mixture of product states. Therefore Table 5 classifies separable states against separable states, and the Discussion's interpretation of the lower accuracy for the 'PPT entangled state' is based on a mislabeled class.
  2. [Section 3.2, Eq. (9), Table 3] The text states both 'full separability for p≤1/3' and 'separable across all bipartitions for p≤1/5' for the three-qubit Werner state; these two statements are mutually inconsistent. Moreover, in tripartite systems the PPT condition is necessary but not sufficient for full separability, so a separability threshold cannot be read off from the PPT eigenvalue. Since the classification methodology is carried over from Sec. 3.1, the labels used for Table 3 appear to treat PPT-entangled states with 1/5 < p ≤ 1/3 as separable, which would contaminate the training data and invalidate the reported accuracy as a measure of entanglement detection.
  3. [Section 3, 'Feature vectors' and 'Computational Efficiency'; Section 2] The central scalability claim is not supported by the proposed pipeline. The feature space is the set of all 4^N−1 non-identity Pauli expectation values, so forming the feature vectors, estimating S_W and S_B, and solving the generalized eigenproblem at O(n^3) with n=4^N−1 all require resources exponential in the qubit number N. Only N=2,3,4 are demonstrated, with no scaling analysis, and the Discussion concedes that 'the high dimensionality of multi-qubit systems also necessitates a large number of measurements.' The paper therefore does not deliver scalable entanglement detection as claimed.
  4. [Section 3.1, Eq. (6), Tables 1, 3, 7] For the Werner-state benchmark, the label is essentially contained in the features. In Eq. (6), the features include ⟨σ_i⊗σ_i⟩ = p s_i, so the separable/entangled boundary p=1/3 is a simple threshold on a single feature. High accuracy on this family is therefore near-tautological and does not demonstrate that FLDA has learned a nontrivial entanglement signature; the same parametric dependence carries over to the GHZ-based three- and four-qubit Werner states.
  5. [Section 3.1, Table 2; Section 4] The concurrence-based two-qubit experiment, which is the non-trivial test, reports a test accuracy of 0.33 in the high-overlap case, below the random-guessing baseline of 0.5, with a Fisher criterion of only 0.012. This directly contradicts the Discussion's statement that 'even in high-overlap scenarios, FLDA often maintains high classification accuracy.' The single most informative numerical result in the paper is therefore evidence against, not for, the method's practical effectiveness.
  6. [Section 3.1, Algorithm 1; Section 4] The evaluation is not reproducible as reported. No code or data are provided; the regularization parameter ϵ in Eq. (5), the normalization/standardization choice, the train/test split, and the procedure for generating the 10,000 samples from N trials = 1000 are not specified; and no error bars are given. In addition, the simulations are noiseless, and the Discussion admits that experimental noise 'would almost certainly lower these accuracy figures.' Since the stated application is real-world experiments, these omissions are load-bearing for the paper's central claim.
minor comments (6)
  1. [Section 3.1, Eqs. (7)-(8)] The matrices are displayed as 3×5 arrays, but the text says W is a 1×15 projection vector; please clarify how the flattening is done.
  2. [Tables 1-7] The quantity 'FLD Threshold' is never defined, and the classification rule in Algorithm 1 (closest projected mean) should be reconciled with the threshold values reported in the tables.
  3. [Section 3.1, Algorithm 1] The relationship between 'N trials = 1000', 'N train = 10000', and the number of features is unclear; please specify how independent samples are generated and whether the features are noisy or exact expectation values.
  4. [Abstract and Introduction] The claims of 'systematically evaluate' and 'high accuracy' are not matched by the four-qubit section, which contains a single high-overlap case.
  5. [References] References [27] and [35] are duplicate entries of the same paper; please remove one.
  6. [Figures 2-8] Axis labels and units are missing, and the text reference 'see Fig. 3, 2' in Section 3.1 is malformed.

Circularity Check

2 steps flagged · score 6.0 of 10

The FLDA method itself is not circular, but the Werner benchmarks are tautological because the label parameter p is encoded directly in the features, and the 'alternative PPT' benchmark leans on a non-independent self-citation; the scalability failure is a separate correctness issue.

  1. self definitional [Sec. 3.1, Eq. (6); same construction in Sec. 3.2 Eq. (9) and Sec. 3.3 Eq. (12), Tables 1, 3, 7]
    "ρ(2) Wer = 1/4 Σ_{i,j=0}^3 c_ij(σ_i⊗σ_j), c_ii = (1 if i=0, p s_i if i=1,2,3) (6) ... Thus, the state is separable for p≤1/3 and entangled for p>1/3. ... Observables like σ_x, σ_y, σ_z and correlations σ_i⊗σ_i are measured ... labeled as separable (p≤ 1/3) or entangled (p > 1/3)."

    The class label is exactly a threshold on the parameter p, while the feature vector supplied to FLDA contains the correlators ⟨σ_i⊗σ_i⟩ = c_ii = p s_i; hence p is recovered as an affine function of a single feature. The optimal linear separator is therefore, by construction, a threshold on that feature, independent of any learned entanglement structure. The high Werner test accuracies (e.g., 1.00 in Table 7) are forced by the data-generation recipe and cannot validate the method. This is self-definitional: the same parameter p fixes both the feature vector and the class label.

  2. self citation load bearing [Sec. 3.2, Eq. (11) and Table 5]
    "Another form of a PPT entangled state for a three-qubit system can be expressed using Pauli operators, as noted in the referenced article: ρ= 1/8 (I⊗I⊗I + I⊗σ_z⊗σ_z + σ_z⊗I⊗σ_z + σ_z⊗σ_z⊗I). ... resulting in a density matrix that is PPT with respect to all possible bipartitions yet exhibits entanglement. [47]"

    The only support for the claim that ρ in Eq. (11) is entangled is reference [47], whose author list includes the present first author. This assertion is load-bearing because Table 5 reports FLDA accuracy for this 'alternative PPT' class. But the operator is diagonal in the computational basis and equals (1/2)(|000⟩⟨000| + |111⟩⟨111|), a convex mixture of product states, so it is separable by construction and the entangled label is not independently established; the self-citation does not supply external evidence.

full rationale

The core FLDA procedure (Sec. 2) is not circular: it takes labeled examples, computes S_B/S_W from the supplied feature vectors, and solves a generalized eigenproblem; it does not assume the answer. The class labels for the main Werner and PPT families come from the Peres–Horodecki/PPT criterion, which is external to FLDA. Where the paper's evidence becomes circular is in the Werner benchmarks: Eq. (6) and its three- and four-qubit analogues write the measured correlators as c_ii = p s_i, so the label-defining parameter p is literally an affine function of a single input feature. A threshold on p is therefore a threshold on that feature, so the high test accuracies in Tables 1, 3, and 7 are guaranteed by construction and cannot validate the method. The concurrence and PPT-state benchmarks are not tautological in the same way, and indeed the concurrence high-overlap case fails (0.33 test accuracy), showing the method is not universally forced. I also flag the 'alternative PPT' state of Eq. (11): its only cited authority is the authors' own ref. [47], and the state is diagonal in the computational basis, hence a separable mixture of product states; the self-citation is load-bearing without independent support. The 'scalable' criticism (exponential growth of the Pauli feature space and O(n^3) eigen-solve) is a correctness/scalability objection, not a circularity, so it does not raise the circularity score. Overall, the central derivation is self-contained, but a substantial part of the reported validation is circular or non-independent.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The paper relies on standard density-matrix and FLDA formalism, but the critical free parameters (epsilon, overlap definitions, feature subsets) are unspecified. The main unsupported domain assumption is that Pauli expectation-value features are sufficient for linear separability, which the paper's own high-overlap results undermine. The PPT entanglement claim for Eq. 11 is unsupported and self-cited.

free parameters (3)
  • epsilon regularization in S_W = unspecified
    Eq. 5 introduces epsilon I regularization for singular within-class scatter, but its value is not given anywhere.
  • overlap level classification = high/medium/low undefined
    Results are reported by 'high', 'medium', and 'low' overlap, but no quantitative definition of these regimes is given.
  • feature set choice = full Pauli basis
    The feature set is stated as Pauli observables, but the precise subset used for two-, three-, and four-qubit states is not specified, and feature selection is acknowledged as arbitrary.
assumptions (3)
  • domain assumption The Peres-Horodecki criterion correctly determines separability for the states under test.
    Used in Sections 3.1 and 3.2 to label training data. It is valid for 2x2 and 2x3 systems, but for three- and four-qubit systems PPT is not sufficient for separability, and the paper does not address this.
  • domain assumption Expectation values of Pauli observables form a sufficient feature space for linear separability of entangled and separable states.
    The paper assumes linear separability in the chosen feature space (Section 3), an assumption its own results contradict in high-overlap cases, where the method performs poorly.
  • ad hoc to paper The alternative PPT state in Eq. 11 is PPT and entangled.
    The paper asserts this without proof and cites the authors' own 2008 paper [47]. The state is a GHZ-diagonal state; its entanglement properties are not established in this text.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scalable Entanglement Detection in Quantum Systems via Fisher Linear Discriminant Analysis." pith.science (2026). https://pith.science/paper/MESPK74S

@misc{pith2026250903233,
  author       = {Pith},
  title        = {Pith review of: Scalable Entanglement Detection in Quantum Systems via Fisher Linear Discriminant Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MESPK74S}},
  note         = {Machine review of arXiv:2509.03233}
}
read the original abstract

Quantum entanglement is the cornerstone of quantum technology and enables quantum devices to outperform classical systems in terms of performance. However, detecting entanglement in high-dimensional systems remains a significant challenge due to the exponential growth of the Hilbert space with the number of particles. In this work, we use machine learning to classify entangled states and separable states, focusing on the application of classical Fisher Linear Discriminant Analysis (FLDA). By adapting classical statistical learning techniques to quantum state discriminant analysis, we present the theoretical foundations, a practical implementation strategy, and the advantages of FLDA in this context. We systematically evaluate the performance of this method on different quantum states and demonstrate its effectiveness as a tool for efficient quantum state classification. Finally, we investigate multi-qubit quantum states with high accuracy and classify these states.

Figures

Figures reproduced from arXiv: 2509.03233 by the authors.

Figure 1
Figure 1. The final state is expressed as |ψ⟩ = cos θ0 2  |00⟩ − isin θ0 2  cos θ1 2  |10⟩ − isin θ0 2  sin θ1 2  |11⟩. For θ0 = π 2 , θ1 = π, the state is maximally entangled (C = 1); for θ0 = 0, θ1 = π, it is separable (C = 0). [27] A two-qubit density matrix is prepared using Werner states or concur￾rence, then measured to extract classical data for FLDA. Observables like σx, σy, σz and correlations σi⊗σi are measured… view at source ↗
Figure 2
Figure 2. Comparison of FLDA plots with different overlaps two-qubit en [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Comparison of FLDA plots with different overlaps two-qubit en [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Comparison of FLDA plots with different overlaps three-qubit [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]
Figure 5
Figure 5. Figure 5: Comparison of FLDA plots with different overlaps three-qubit [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: Comparison of FLDA plots with different overlaps three-qubit PPT [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 7
Figure 7. Figure 7: FLDA plots with high overlap for three-qubit biseparable States [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: FLDA plots with high overlap for four-qubit Werner State and [PITH_FULL_IMAGE:figures/full_fig_p020_8.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 42 canonical work pages

  1. [1]

    Can quantum-mechanical de- scription of physical reality be considered complete?,

    A. Einstein, B. Podolsky, and N. Rosen, “Can quantum-mechanical de- scription of physical reality be considered complete?,”Physical Review, vol. 47, no. 10, pp. 777–780, 1935. 21

    work page 1935

  2. [2]

    On the einstein podolsky rosen paradox,

    J. S. Bell, “On the einstein podolsky rosen paradox,”Physics Physique Fizika, vol. 1, no. 3, pp. 195–200, 1964

    work page 1964

  3. [3]

    Quantum entanglement,

    R. Horodecki, P. Horodecki, M. Horodecki, and K. Horodecki, “Quantum entanglement,”Reviews of Modern Physics, vol. 81, no. 2, pp. 865–942, 2009

    work page 2009

  4. [4]

    M. A. Nielsen and I. L. Chuang,Quantum Computation and Quantum Information. Cambridge University Press, 10th anniversary edition ed., 2010

    2010

  5. [5]

    Quantum computing in the nisq era and beyond,

    J. Preskill, “Quantum computing in the nisq era and beyond,”Quantum, vol. 2, p. 79, 2018

    2018

  6. [6]

    Quantum computing with realistically noisy devices,

    E. Knill, “Quantum computing with realistically noisy devices,”Nature, vol. 434, no. 7029, pp. 39–44, 2005

    work page 2005

  7. [7]

    Teleporting an unknown quantum state via dual classical and einstein-podolsky-rosen channels,

    C. H. Bennett, G. Brassard, C. Crépeau, R. Jozsa, A. Peres, and W. K. Wootters, “Teleporting an unknown quantum state via dual classical and einstein-podolsky-rosen channels,”Physical Review Letters, vol. 70, no. 13, pp. 1895–1899, 1993

    work page 1993

  8. [8]

    Multiparticle entanglement and its applications to quan- tum information science,

    J.-W. Pan, Z.-B. Chen, C.-Y. Lu, H. Weinfurter, A. Zeilinger, and M. Żukowski, “Multiparticle entanglement and its applications to quan- tum information science,”Reviews of Modern Physics, vol. 84, no. 2, pp. 777–838, 2012

    work page 2012

Show all 47 references
  1. [9]

    Ad- vances in quantum cryptography,

    S.Pirandola, U.L.Andersen, L.Banchi, M.Berta, D.Bunandar, R.Col- beck, D. Englund, T. Gehring, R. H. Hadfield, H. Hübel,et al., “Ad- vances in quantum cryptography,”Advances in Optics and Photonics, vol. 12, no. 4, pp. 1012–1236, 2020

    2020

  2. [10]

    Quantum sensing,

    C. L. Degen, F. Reinhard, and P. Cappellaro, “Quantum sensing,”Re- views of Modern Physics, vol. 89, no. 3, p. 035002, 2017

    2017

  3. [11]

    Entanglement detection,

    O. Gühne and G. Tóth, “Entanglement detection,”Physics Reports, vol. 474, no. 1-6, pp. 1–75, 2009

    2009

  4. [12]

    Advances in quantum metrology,

    V. Giovannetti, S. Lloyd, and L. Maccone, “Advances in quantum metrology,”Nature Photonics, vol. 5, no. 4, pp. 222–229, 2011. 22

    2011

  5. [13]

    Entanglement cer- tification from theory to experiment,

    N. Friis, G. Vitagliano, M. Malik, and M. Huber, “Entanglement cer- tification from theory to experiment,”Nature Reviews Physics, vol. 1, no. 1, pp. 72–87, 2019

    2019

  6. [14]

    Entanglement detection and quantification in high-dimensional systems,

    H. J. Briegel, G. Tóth, and O. Gühne, “Entanglement detection and quantification in high-dimensional systems,”Physical Review Letters, vol. 127, no. 14, p. 140502, 2021

    2021

  7. [15]

    Quantum entanglement detection with machine learning,

    S. Lu, J. Chen, Y. Zhang, Z. Li, and B. Zeng, “Quantum entanglement detection with machine learning,”Physical Review Letters, vol. 125, no. 20, p. 200501, 2020

    2020

  8. [16]

    Efficient entanglement detection in few-measurement scenarios,

    J. Shang, Z. Zhang, and O. Gühne, “Efficient entanglement detection in few-measurement scenarios,”Physical Review A, vol. 104, no. 5, p. 052412, 2021

    2021

  9. [17]

    High-dimensional entanglement certification with minimal measurements,

    Z. Li, X. Wang, Y. Ma, and H. Zhang, “High-dimensional entanglement certification with minimal measurements,”npj Quantum Information, vol. 8, no. 1, p. 92, 2022

    2022

  10. [18]

    Scalable multipartite entan- glement detection using randomized measurements,

    Y. Zhang, D. Zhou, S. Chen, and B. Zeng, “Scalable multipartite entan- glement detection using randomized measurements,”Quantum, vol. 7, p. 1056, 2023

    2023

  11. [19]

    Adaptive protocols for high-dimensional entanglement detection,

    H. Huang, J. Chen, and D. Lu, “Adaptive protocols for high-dimensional entanglement detection,”Physical Review A, vol. 110, no. 2, p. 022401, 2024

    2024

  12. [20]

    Efficient entanglement certification for high-dimensional systems,

    S. Chen, D. Zhou, and B. Zeng, “Efficient entanglement certification for high-dimensional systems,”npj Quantum Information, vol. 9, no. 1, p. 45, 2023

    2023

  13. [21]

    Multipartite entanglement de- tection with minimal resources,

    X. Wang, Z. Li, H. Zhang, and Y. Ma, “Multipartite entanglement de- tection with minimal resources,”Physical Review A, vol. 109, no. 4, p. 042401, 2024

    2024

  14. [22]

    Advances in quantum entangle- ment detection: Beyond traditional witnesses,

    Y.-J. Kim, K.-H. Lee, and C. Kim, “Advances in quantum entangle- ment detection: Beyond traditional witnesses,”Quantum Science and Technology, vol. 8, no. 2, p. 025005, 2023. 23

    2023

  15. [23]

    Efficient quantum state tomography for high-dimensional systems,

    Y. Yang, H. Zhang, and X. Wang, “Efficient quantum state tomography for high-dimensional systems,”Quantum Science and Technology, vol. 7, no. 3, p. 035015, 2022

    2022

  16. [24]

    Bell inequalities tailored to high-dimensional systems,

    R. Augusiak, J. Tura, and M. Lewenstein, “Bell inequalities tailored to high-dimensional systems,”Physical Review A, vol. 89, no. 5, p. 052303, 2014

    2014

  17. [25]

    Bell nonlocality,

    N. Brunner, D. Cavalcanti, S. Pironio, V. Scarani, and S. Wehner, “Bell nonlocality,”Reviews of Modern Physics, vol. 86, no. 2, pp. 419–478, 2014

    2014

  18. [26]

    Scalable entanglement detection via machine learning,

    S. Yu, Y. Zhang, J. Chen, D. Lu, and B. Zeng, “Scalable entanglement detection via machine learning,”Physical Review A, vol. 106, no. 3, p. 032401, 2022

    2022

  19. [27]

    Entanglement detection with quantum support vector machine (qsvm) on near-term quantum devices,

    M. Mahdian and Z. Mousavi, “Entanglement detection with quantum support vector machine (qsvm) on near-term quantum devices,”Scien- tific Reports, vol. 15, no. 1, pp. 1–15, 2025

    2025

  20. [28]

    Optimal entanglement witness of multi- partite systems using support vector machine approach,

    M. Mahdian and Z. Mousavi, “Optimal entanglement witness of multi- partite systems using support vector machine approach,”arXiv preprint arXiv:2504.18163, 2025

    2025 arXiv

  21. [29]

    Machine-learning-enhanced entanglement detection under noisy quan- tum measurements,

    M. Mahdian, A. Babapour-Azar, Z. Mousavi, and R. Khanjani-Shiraz, “Machine-learning-enhanced entanglement detection under noisy quan- tum measurements,”arXiv preprint arXiv:2507.05476, 2025

    2025

  22. [30]

    Machine learning and the physical sciences,

    G. Carleo, I. Cirac, K. Cranmer, L. Daudet, M. Schuld, N. Tishby, L. Vogt-Maranto, and L. Zdeborová, “Machine learning and the physical sciences,”Reviews of Modern Physics, vol. 91, no. 4, p. 045002, 2019

    2019

  23. [31]

    Reformulation of the no-free-lunch theorem for entangled datasets,

    K. Sharma, M. Cerezo, Z. Holmes, L. Cincio, A. S. Sornborger, and P. J. Coles, “Reformulation of the no-free-lunch theorem for entangled datasets,”Physical Review Letters, vol. 128, no. 7, p. 070501, 2022

    2022

  24. [32]

    Direct entanglement detection of quantum systems using machine learning,

    Y. Huang, L. Che, C. Wei,et al., “Direct entanglement detection of quantum systems using machine learning,”npj Quantum Information, vol. 11, p. 29, 2025. 24

    2025

  25. [33]

    Separability-entanglementclassifierviamachinelearning,

    S. Lu, S. Huang, K. Li, J. Li, J. Chen, D. Lu, Z. Ji, Y. Shen, D. Zhou, andB.Zeng, “Separability-entanglementclassifierviamachinelearning,” Physical Review A, vol. 98, no. 1, p. 012315, 2018

    2018

  26. [34]

    Machine learning for quantum state classification,

    J. Gray, L. Banchi, and P. Kok, “Machine learning for quantum state classification,”Physical Review A, vol. 98, no. 5, p. 052306, 2018

    2018

  27. [35]

    Entanglement detection with quantum support vector machine (qsvm) on near-term quantum devices,

    M. Mahdian and Z. Mousavi, “Entanglement detection with quantum support vector machine (qsvm) on near-term quantum devices,”Scien- tific Reports, vol. 15, p. 11931, 2025

    2025

  28. [36]

    Interpretable machine learn- ing for entanglement classification,

    J. Chen, Z. Li, H. Zhang, and X. Wang, “Interpretable machine learn- ing for entanglement classification,”Physical Review A, vol. 108, no. 4, p. 042402, 2023

    2023

  29. [37]

    Unsupervised learn- ing for entanglement detection,

    Y. Wang, Y. Li, S. Zhang, D. Zhou, and B. Zeng, “Unsupervised learn- ing for entanglement detection,”Physical Review A, vol. 102, no. 6, p. 062406, 2020

    2020

  30. [38]

    Deep learning of quantum entanglement from in- complete measurements,

    D. Koutnýet al., “Deep learning of quantum entanglement from in- complete measurements,”Science Advances, vol. 9, no. 29, p. eadg1719, 2023

    2023

  31. [39]

    Efficient unsupervised entan- glement detection with neural networks,

    Y. Li, Y. Wang, D. Zhou, and B. Zeng, “Efficient unsupervised entan- glement detection with neural networks,”npj Quantum Information, vol. 10, no. 1, p. 38, 2024

    2024

  32. [40]

    Detecting entanglement in high-spin quan- tum systems via a stacking ensemble of machine learning models,

    M. Y. Abd-Rabbouet al., “Detecting entanglement in high-spin quan- tum systems via a stacking ensemble of machine learning models,”arXiv preprint arXiv:2507.12775, 2025

    2025 arXiv

  33. [41]

    Machine learning and quantum physics,

    V. Dunjko and H. J. Briegel, “Machine learning and quantum physics,” Reports on Progress in Physics, vol. 81, no. 7, p. 074001, 2018

    2018

  34. [42]

    Quantum machine learning,

    J. Biamonte, P. Wittek, N. Pancotti, P. Rebentrost, N. Wiebe, and S. Lloyd, “Quantum machine learning,”Nature, vol. 549, no. 7671, pp. 195–202, 2017

    2017

  35. [43]

    The use of multiple measurements in taxonomic prob- lems,

    R. A. Fisher, “The use of multiple measurements in taxonomic prob- lems,”Annals of Eugenics, vol. 7, no. 2, pp. 179–188, 1936. 25

    1936

  36. [44]

    Hastie, R

    T. Hastie, R. Tibshirani, and J. Friedman,The Elements of Statistical Learning: Data Mining, Inference, and Prediction. Springer, 2nd ed., 2009

    2009

  37. [45]

    C. M. Bishop and N. M. Nasrabadi,Pattern recognition and machine learning, vol. 4. Springer, 2006

    2006

  38. [46]

    Classification of mixed three-qubit states,

    A. Acín, D. Bruß, M. Lewenstein, and A. Sanpera, “Classification of mixed three-qubit states,”Physical Review Letters, vol. 87, no. 4, p. 040401, 2001

    2001

  39. [47]

    Detect- ing some three-qubit mub diagonal entangled states via nonlinear opti- mal entanglement witnesses,

    M. Jafarizadeh, M. Mahdian, A. Heshmati, and K. Aghayar, “Detect- ing some three-qubit mub diagonal entangled states via nonlinear opti- mal entanglement witnesses,”The European Physical Journal D, vol. 50, pp. 107–121, 2008. 26

    2008

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.